[ https://issues.apache.org/jira/browse/LUCENE-2878?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12990217#comment-12990217 ]
Simon Willnauer commented on LUCENE-2878: ----------------------------------------- {quote} How does/should scoring work? EG do the SpanQueries score according to the details of which position intervals match? {quote} I didn't pay any attention to scoring yet. IMO scoring should be left to the query which is using the positions so a higher level query like a NearQuery could just put a custom scorer on top of a boolean conjunction and apply its own proximity based score. This should be done after we have the infrastructure to do it. I think that opens up some nice scoring improvements. I am not sure if we should add proximity scoring to existing queries, I rather aim towards making it easy to customize. {quote} The part I'm wondering about is what API we should use for communicating positions of the sub scorers in a BooleanQuery to consumers like position filters (for matching) or eg Highlighter (which really should be a core functionality that works w/ any query). Multiplying out ("denormalizing") all combinations (into a flat stream of PositionIntervals) is going to be too costly in general, I think? {quote} I thought about that for a while and I think we should enrich the PosIntervalIterator API to enable the caller to pull the actual subintervals instead of an Interval from the next method. Something like this: {code} public abstract class PositionIntervalIterator implements Serializable{ public abstract PositionInterval next() throws IOException; /** *Returns all sub interval for the next accepted interval. **/ public abstract PositionIntervalIterator nextSubIntervals() throws IOException; public abstract PositionIntervalIterator[] subs(boolean inOrder); {code} so that if you are interested in the plain positions for eventually each term like highlighting you can pull them per match occurence. That way you have positional matching and you can iterate the subs. {quote} Maybe, instead of the denormalized stream, we could present a UnionPositionsIntervalIterator, which has multiple subs, where each sub is its own PositionIntervalIterator? This way eg a NEAR query could filter these subs in parallel (like a merge sort) looking for a match, and (I think) then presenting its own union iterator to whoever consumes it? Ie it'd only let through those positions of each sub that satisfied the NEAR constraint. {quote} I don't get that entirely ;) bq. Does it make sense that we could just want AttributeSources as we go here? you mean like we are not extending Scorer but add an AttributeSource to it? I think this is really a core API and should be supported directly > Allow Scorer to expose positions and payloads aka. nuke spans > -------------------------------------------------------------- > > Key: LUCENE-2878 > URL: https://issues.apache.org/jira/browse/LUCENE-2878 > Project: Lucene - Java > Issue Type: Improvement > Components: Search > Affects Versions: Bulk Postings branch > Reporter: Simon Willnauer > Assignee: Simon Willnauer > Attachments: LUCENE-2878.patch, LUCENE-2878.patch, LUCENE-2878.patch, > LUCENE-2878.patch > > > Currently we have two somewhat separate types of queries, the one which can > make use of positions (mainly spans) and payloads (spans). Yet Span*Query > doesn't really do scoring comparable to what other queries do and at the end > of the day they are duplicating lot of code all over lucene. Span*Queries are > also limited to other Span*Query instances such that you can not use a > TermQuery or a BooleanQuery with SpanNear or anthing like that. > Beside of the Span*Query limitation other queries lacking a quiet interesting > feature since they can not score based on term proximity since scores doesn't > expose any positional information. All those problems bugged me for a while > now so I stared working on that using the bulkpostings API. I would have done > that first cut on trunk but TermScorer is working on BlockReader that do not > expose positions while the one in this branch does. I started adding a new > Positions class which users can pull from a scorer, to prevent unnecessary > positions enums I added ScorerContext#needsPositions and eventually > Scorere#needsPayloads to create the corresponding enum on demand. Yet, > currently only TermQuery / TermScorer implements this API and other simply > return null instead. > To show that the API really works and our BulkPostings work fine too with > positions I cut over TermSpanQuery to use a TermScorer under the hood and > nuked TermSpans entirely. A nice sideeffect of this was that the Position > BulkReading implementation got some exercise which now :) work all with > positions while Payloads for bulkreading are kind of experimental in the > patch and those only work with Standard codec. > So all spans now work on top of TermScorer ( I truly hate spans since today ) > including the ones that need Payloads (StandardCodec ONLY)!! I didn't bother > to implement the other codecs yet since I want to get feedback on the API and > on this first cut before I go one with it. I will upload the corresponding > patch in a minute. > I also had to cut over SpanQuery.getSpans(IR) to > SpanQuery.getSpans(AtomicReaderContext) which I should probably do on trunk > first but after that pain today I need a break first :). > The patch passes all core tests > (org.apache.lucene.search.highlight.HighlighterTest still fails but I didn't > look into the MemoryIndex BulkPostings API yet) -- This message is automatically generated by JIRA. - For more information on JIRA, see: http://www.atlassian.com/software/jira --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org